EP0677820A2 - Magnifying digital image using edge mapping - Google Patents
Magnifying digital image using edge mapping Download PDFInfo
- Publication number
- EP0677820A2 EP0677820A2 EP94119855A EP94119855A EP0677820A2 EP 0677820 A2 EP0677820 A2 EP 0677820A2 EP 94119855 A EP94119855 A EP 94119855A EP 94119855 A EP94119855 A EP 94119855A EP 0677820 A2 EP0677820 A2 EP 0677820A2
- Authority
- EP
- European Patent Office
- Prior art keywords
- pixels
- image
- pixel
- locations
- original
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/403—Edge-driven scaling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4023—Decimation- or insertion-based scaling, e.g. pixel or line decimation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/12—Edge-based segmentation
Definitions
- the present invention relates generally to magnifying a digital image and more particularly to magnifying a digital image using an edge map.
- Figure 1 shows an image before magnification
- Figure 2 shows the image after magnification by such a method. Observe the small square blocks of unpleasant pixels about 0.5 mm wide making up the image, and the jagged edges, such as the edges of the petals of the flowers.
- FIG. 3 shows a magnified image formed by such a bilinear-interpolation method. This method substantially removes the blocks of unpleasant pixels and the jagged edges. However, the crisp edges in the original image are blurred due to the averaging effect.
- the present invention provides methods to magnify a digital image that substantially maintains its crisp edges.
- Prior arts tend to generate small blocks of unpleasant pixels with jagged edges or blur the crisp edges of the original digital image.
- the digital image has many original pixels. Each pixel has a level, and the locations of the pixels have a resolution.
- an edge map is produced from the digital image.
- the map with its many boundaries are projected onto the digital image to help generate the levels of one or more additional pixels in locations among the original pixels.
- the levels of one or more original pixels are manipulated without crossing any boundaries set by the edge map. After the additional pixels are generated, the distances between all the pixels are expanded so that the resolution of the locations of the pixels becomes the same as the original pixels. Thus, the digital image is magnified.
- the original image is enlarged, before producing the edge map, to the size of the magnified image by expanding proportionally the distances separating the original pixels. Then the edge map is produced and projected onto the enlarged image to generate additional pixels. After the step of generating, the resolution of the locations of the pixels becomes the same as the original pixels.
- the magnified image is substantially as crisp as the original image.
- Figure 1 shows an image before magnification.
- Figure 2 shows the image magnified by a prior arts pixel-replication technique.
- Figure 3 shows the image magnified by a prior arts bilinear-interpolation technique.
- Figure 4 shows one preferred method to magnify a digital image in the present invention.
- Figure 5 shows one preferred method to pick locations with significant changes in levels.
- Figure 6 shows one preferred method to pick locations of zero-crossings.
- Figure 7 shows a boundary with original pixels and temporary pixels for the present invention.
- Figure 8 shows a boundary with original pixels, temporary pixels and target pixels for the present invention.
- Figure 9 shows another preferred method to pick locations with significant changes in levels.
- Figure 10 shows an example of an edge map in the present invention.
- FIGS 11A-E show two examples of interpolation methods in the present invention.
- Figures 12A-B show a method to correct for abnormalities next to boundaries.
- Figure 13 shows a magnified image generated using an edge map.
- Figure 14 shows another preferred method to magnify a digital image in the present invention.
- Figure 15 shows the feedback method in the present invention.
- Figure 16 shows a two-dimensional example of a prior art averaging process.
- Figure 17 shows an example of the step of modifying in the present invention.
- Figure 18 shows a magnified image generated with the feedback method.
- Figure 4 shows one preferred method 100 to magnify a digital image with many original pixels.
- the pixels are arranged with a certain resolution, for example, 300 dots or pixels per inch.
- Each pixel has a level, which describes the intensity at that pixel.
- the intensity level may vary from 0 to 255.
- a non-continuous tone image such as a halftone image, preferably, it is transformed to a continuous tone image before it is magnified; this process will be further described later in the detailed description.
- any region with a crisp-edge is a region with a significant change in level.
- the first step in the preferred embodiment 100 is to analyze, 102, the image to pick all the locations with significant change in levels. This can be done visually or analytically.
- FIG. 5 shows one preferred analytical method, 200, to pick locations with significant changes in levels.
- Each pixel is represented, 202, by an impulse function with its magnitude being the level at that pixel.
- select, 204 a singly-peaked continuous function.
- this can be a normalized Gaussian function, having a second standard deviation covering the width of 3 original pixels.
- the selected function is a two-dimensional surface.
- the selected continuous function is differentiated, 206, twice. If the function is a Gaussian function, then the second derivative is known as the Laplacian of the Gaussian.
- the result from the differentiation step is convolved, 208, with all the impulse functions, and the outputs are superimposed spatially to generate a convolved result.
- Pick, 212 all the locations that have substantially zero value from the convolved result. All the crisp-edge locations in the image before magnification correspond to locations in the convolved results with substantially zero values.
- Figure 6 shows one preferred method, 250, to pick locations that have substantially zero values or to pick the "zero-crossings" from the convolved results. There are a few ways to select the "zero-crossings.” One can solve analytically the convolved result to find all the locations having zeros. This calculation might be quite cumbersome. Another preferred way is to solve the convolved result at selected positions, as shown in Figure 6, in combination with Figure 7.
- N is 2
- D is 3.3 mils for an image with a 300 dots/inch resolution.
- picking, 250, locations with substantially zero value first, one forms, 252, temporary pixels at positions offset from the original pixels, with the minimum distance between adjacent temporary pixels being D/N.
- the crosses represent the locations of the original pixels at 300 dots per inch; the circles represent the temporary pixels; and the line marked EDGE represents the theoretical locations with significant changes in levels.
- the minimum distance between adjacent circles in Figure 7 is D/N or 1.67 mils, which gives a resolution of 600 dots per inch.
- the circles are offset from the crosses.
- the amount of offset is D/(2N) in both the X and the Y direction, with N being 2.
- the convolved result is evaluated, 254, at the temporary pixels.
- the levels of adjacent temporary pixels are compared, 256, to select, 258, those pairs of adjacent temporary pixels whose levels change sign. For example, there will be a sign change between the levels at circles 272 and 274, and between the levels at circles 274 and 276. Note that in the present application, changing sign includes changing from zero to either a positive or negative value.
- the method 250 may be used in combination with Figure 8 as another scheme to pick locations that have substantially zero values or to pick the "zero-crossings" from the convolved result.
- This scheme reduces the amount of computation required and is especially suitable for even number magnification higher than 2.
- the locations of the temporary pixels have a resolution doubling the original resolution, with some of them not displaced from the original pixels.
- the locations of the target pixels have a resolution four times the original resolution, and are displaced from the temporary pixels by D/(2N) in both the X and the Y direction, with N being 4 for the example shown in Figure 8.
- the levels are again evaluated, 254, at the temporary pixels, with adjacent temporary pixels compared, 256, to select, 258, those pairs with sign change in their levels.
- the signs of the levels of the temporary pixels 282 and 283 are different, with the level of 283 being closer to zero.
- the target pixel 288 is chosen.
- Figure 9 shows another preferred embodiment 300 to pick locations with significant changes in levels.
- Each pixel is again represented, 302, by an impulse function with its magnitude being the level at that pixel.
- select, 304 a singly-peaked continuous function, such as the same Gaussian function picked in the previous embodiment 200.
- the selected function is convolved, 306, with all the impulse functions, and the outputs are superimposed together to get a convolved result.
- the convolved result is differentiated, 310, twice to get a differentiated result.
- pick, 312 all the locations that have substantially zero from the differentiated result. All the crisp-edge locations have values that are substantially zero.
- threshold, 214 or 314 means removing some of the picked locations if the local level variances of the picked locations are less than a pre-selected value.
- the pre-selected value should be small because it is acceptable to have more locations which are perceived to be zero-crossings. However, missing zero-crossings would not be preferable. In one example, the pre-selected value is 0.03.
- the locations should be connected, 104, to form boundaries for an edge map. However not all the locations should be connected. In one preferred embodiment, one connects any two picked locations to form a boundary if the distance between the locations is less than or equal to 1.414 (or the square root of 2) times the minimum distance between two adjacent original pixels. For a digital image with 300 dots per inch, the minimum distance is 3.3 mils, and 1.414 times of 3.3 is equal to 4.7 mils.
- Figure 10 shows an example of a two-dimensional edge map formed by the above described process on the original image shown in Figure 1. That map is enlarged to the size of the desired magnified image.
- the edge map is formed, it is projected, 106, onto the digital image. Then one generates, 108, many additional pixels in locations among the original pixels by manipulating the original pixels without crossing any boundaries set by the edge map.
- the number of additional pixels formed depends on the magnification desired. If one wants M times magnification in one dimension and N times in another dimension, one needs to increase the number of pixels in the image to M times N, which means forming additional number of pixels in appropriate locations.
- Figures 11A-D illustrate the first example, which is especially suitable for even numbers of magnifications.
- Crosses represent the locations of original pixels, for example, 379 and 383.
- Triangles represent the locations of additional pixels, for example, 385.
- calculating the level of the additional pixel 385 one interpolates the levels of the original pixels, 377, 379, 381; but one does not use the level of the original pixel 383 because that would imply crossing a boundary, EDGE 1, in the edge map.
- Figure 11B shows an example of finding the level at additional pixel 385, based on the levels at three original pixels 377, 379 and 381.
- the fictitious pixel 382 is located at the location of the original pixel 383. From the levels of pixels 377, 379, 381 and 382, one calculates the level of the additional pixel 385 using bilinear interpolation. The above calculations use proportionality based on distance and will not be further discussed here.
- Figure 11C shows examples of finding the level at the additional pixel 392, based on two original pixels, 383, 387.
- Two situations will be described, the first situation involving just one boundary next to the original pixels, and the second situation involving the original pixels surrounded by two boundaries.
- one preferred method is to extrapolate based on the neighbors of the original pixels.
- the value at the fictitious pixel 390 is extrapolated from the levels in the original pixels 381 and 383, while the value at the fictitious pixel 391 is extrapolated from the levels in the original pixels 389 and 387.
- one preferred method is to copy the value of the level in the original pixels 383 and 387 to the fictitious pixels 390 and 391 respectively.
- the locations of the fictitious pixels 390 and 391 are the locations of the original pixels 388 and 386 respectively. From the levels of pixels 383, 390, 387 and 391, one calculates the level of the additional pixel 392 using bilinear interpolation.
- Figure 11D shows an example of finding the level at the additional pixel 394, based on one original pixel, 381.
- One method is to duplicate the level of the original pixel, 381, at the locations of the fictitious pixels, 393, 395 and 397, which are the locations of the original pixels 383, 396 and 398 respectively. From the levels of the pixels 381, 393, 395 and 397, one calculates the level of the additional pixel 394 using bilinear interpolation.
- Figure 11E illustrates the second example to find the levels at additional levels, which is especially suitable for odd numbers of magnifications.
- the main difference between this example and the first example is that some additional pixels are located at the original pixel locations.
- the methods to generate the levels of the additional pixels can be similar to the methods as described in Figures 11B-D.
- a size of a window 475 is selected, 452.
- One example is a size that covers 2 by 2 original pixels.
- center, 454 the window at every additional pixel close to a found boundary. For example, center the window 475 at the additional pixel 477.
- the average level of the original pixels on the side of the boundary of the additional pixel within the window For example, calculate the average level of the original pixels, 479, 481 and 483. If the average level differs by more than a pre-set value from the level of the additional pixel, replace, 458, the level of the additional pixel with the average value.
- the pre-set value is one standard deviation of the levels of the original pixels. This process is performed on every pixel in close proximity to a boundary. If an additional pixel is bounded by two boundaries, then the original pixels averaged are only those within the window on the side of the additional pixel and within the two boundaries.
- the distances between the pixels are expanded, 112, so that the resolution of the locations of the pixels are the same as the original pixels.
- all the crosses in the digital image in Figure 11A has a resolution of 300 pixels per inch.
- the triangles have a resolution of 600 pixels per inch.
- the size of the image is then expanded by expanding the distances between the additional pixels so that the resolution of the locations of the additional pixels goes back to 300 pixels per inch. This means expanding the image two times in each dimension.
- Figure 13 shows the magnified digital image based on the above described method.
- the magnified image is more crisp than the image in Figure 3, for example, as in the stamens of the off-white lily on the right hand side of the Figure or as in the gaps between flowers.
- Figure 14 shows a modified preferred embodiment 498, in which the distances between pixels are expanded in the beginning.
- the original digital image is enlarged, 490 by expanding proportionally the distances between the original pixels.
- an edge map is produced, 492, by one of the preferred methods described above.
- the produced edge map is projected, 494, onto the enlarged image to generate, 496, additional pixels by methods as described above.
- the resolution of the locations of the pixels becomes the same as the original pixels, and the digital image is magnified.
- the image in Figure 13 is further improved by a feedback method, 500, shown in Figure 15.
- the original digital image is duplicated, 502, to form a duplicated image.
- the duplicated image is then magnified, 504, R times.
- This magnification process may be based on the edge map method as described above or may be based on other means. However, the magnification process is preferably not the pixel replication process, which generates the image shown in Figure 2.
- the magnified image is then de-magnified, 506, R times.
- the demagnification process is based on averaging. Every fixed number of pixels are replaced with one pixel whose level is the average level of the fixed number of pixels.
- Figure 16 shows a two-dimensional example of the averaging process for a two-times demagnification. Every four pixel is replaced by one pixel; for example, the pixels 550, 552, 554 and 556 in the magnified image 570 is replaced by the pixel 558 in the de-magnified image 572.
- the level of the pixel 558 is equal to the average of the levels of the four pixels in the magnified image 570.
- the location of the pixel 558 is in the center of the four pixels 550, 552, 554 and 556.
- the de-magnifying step is preferably not the inverse of the magnifying step. In other words, the magnification step is preferably not a pixel-replication process.
- the difference between the de-magnified image and the original digital image at every pixel is a pixel-to-pixel comparison method.
- the duplicated image is modified, 512.
- This modifying step includes modifying every pixel of the duplicated image by a selected multiple of the difference found at that pixel from the step of calculation, 508.
- Figure 16 shows an example of the step of modifying.
- the original image has a pixel 600 with a grey-level of 200. That pixel is compared to the pixel, 602, at the corresponding position in the de-magnified image.
- the pixel 602 has a grey-level of 240.
- the difference value is 40.
- a multiple of the difference value is used to modify the corresponding pixel, 604, in the duplicated image. If the multiple is too large, one may get an "oscillation.” If the multiple is too small, one may take a long time before one gets satisfactory results. The multiple value can be found experimentally. Note also that the three pixels, 600, 602 and 604, are all in substantially the same location for the pixel-to-pixel operation described.
- Figure 18 shows an example of applying the feedback method on the image shown in Figure 13, with the pre-selected value set as 2.5 and the multiple value set as 1.
- the multiple value preferably should be between 0 and 2.
- the magnified image formed after the feedback method has significantly better level consistency. Also, many details in the original image are maintained in the magnified image; for example, compare the veins on the petals of the flowers in Figure 13 and Figure 17.
- This feedback process is applicable not just to two-dimensional images. It is equally applicable for magnifying images with other dimensions, such as one or three dimensional images.
- the difference is the fixed number of pixels to be averaged.
- the fixed number is R.
- the fixed number is R * R * R.
- the above preferred embodiment describes the magnification factors being equal for all the dimensions.
- the magnification factors are different for different dimensions.
- the two dimensional magnification may be R * N, with R not equal to N.
- the present invention is not just applicable to continuous tone images. It can also be applied to non-continuous tone images, such as halftone images.
- non-continuous tone images such as halftone images.
- halftone images the images are first changed to continuous tones, based on inverse halftoning techniques. Then images are magnified as described above. After the magnification process, the images are halftoned to produce halftone images. Inverse halftoning and halftoning techniques are obvious to those with ordinary skill in the arts and will not be further described in the present disclosure.
- an edge map significantly maintains the crisp-edge features of a magnified image, and the feedback method further improves the quality of the magnified image.
- the invention is applicable to many other areas than those just described.
- the invention is applicable to three-dimensional images, such as a holographic image; in that case, the edge map would not be just lines, but surfaces.
- Another example is to apply the invention to color images.
- the level of each pixel is the intensity at that pixel.
- the level of each pixel can be identified by the intensities of their primaries, such as red, green and blue; the present invention is applied to each color individually, and then the results of the three colors are combined at each pixel to re-generate the color images.
- Another preferred embodiment for color images is to identify the level of each pixel as its luminous at that pixel. Then the present invention is applied to the luminous value as described above.
- This invention can be applied to one-dimensional images.
- the image is a one-dimensional waveform or a signal, like a voice signal.
- Another area of application is in video.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
- Editing Of Facsimile Originals (AREA)
- Controls And Circuits For Display Device (AREA)
Abstract
Description
- The present invention relates generally to magnifying a digital image and more particularly to magnifying a digital image using an edge map.
- The proliferation of digital imagery cries for better ways to magnify an image. In a traditional magnification process, additional pixels are added into the original pixels. Then the size of the image is magnified so that the distance between adjacent pixels is maintained to be the same as that in the original digital image.
- Different methods are available to add the additional pixels. One method simply replicates pixels--adds a number of pixels surrounding every existing pixel to form a block of pixels with the same level. Figure 1 shows an image before magnification and Figure 2 shows the image after magnification by such a method. Observe the small square blocks of unpleasant pixels about 0.5 mm wide making up the image, and the jagged edges, such as the edges of the petals of the flowers.
- Another method generates the additional pixels by interpolating the levels of adjacent original pixels. Figure 3 shows a magnified image formed by such a bilinear-interpolation method. This method substantially removes the blocks of unpleasant pixels and the jagged edges. However, the crisp edges in the original image are blurred due to the averaging effect.
- It should be apparent from the foregoing that there is still a need for a method to magnify a digital image that maintains its crisp edges.
- The present invention provides methods to magnify a digital image that substantially maintains its crisp edges. Prior arts tend to generate small blocks of unpleasant pixels with jagged edges or blur the crisp edges of the original digital image.
- The digital image has many original pixels. Each pixel has a level, and the locations of the pixels have a resolution. In one preferred embodiment of the present invention, an edge map is produced from the digital image. The map with its many boundaries are projected onto the digital image to help generate the levels of one or more additional pixels in locations among the original pixels. In the process to generate additional pixels, the levels of one or more original pixels are manipulated without crossing any boundaries set by the edge map. After the additional pixels are generated, the distances between all the pixels are expanded so that the resolution of the locations of the pixels becomes the same as the original pixels. Thus, the digital image is magnified.
- In another preferred embodiment, the original image is enlarged, before producing the edge map, to the size of the magnified image by expanding proportionally the distances separating the original pixels. Then the edge map is produced and projected onto the enlarged image to generate additional pixels. After the step of generating, the resolution of the locations of the pixels becomes the same as the original pixels.
- With the edge map, the magnified image is substantially as crisp as the original image.
- Other aspects and advantages of the present invention will become apparent from the following detailed description, which, when taken in conjunction with the accompanying drawings, illustrates by way of example the principles of the invention.
- Figure 1 shows an image before magnification.
- Figure 2 shows the image magnified by a prior arts pixel-replication technique.
- Figure 3 shows the image magnified by a prior arts bilinear-interpolation technique.
- Figure 4 shows one preferred method to magnify a digital image in the present invention.
- Figure 5 shows one preferred method to pick locations with significant changes in levels.
- Figure 6 shows one preferred method to pick locations of zero-crossings.
- Figure 7 shows a boundary with original pixels and temporary pixels for the present invention.
- Figure 8 shows a boundary with original pixels, temporary pixels and target pixels for the present invention.
- Figure 9 shows another preferred method to pick locations with significant changes in levels.
- Figure 10 shows an example of an edge map in the present invention.
- Figures 11A-E show two examples of interpolation methods in the present invention.
- Figures 12A-B show a method to correct for abnormalities next to boundaries.
- Figure 13 shows a magnified image generated using an edge map.
- Figure 14 shows another preferred method to magnify a digital image in the present invention.
- Figure 15 shows the feedback method in the present invention.
- Figure 16 shows a two-dimensional example of a prior art averaging process.
- Figure 17 shows an example of the step of modifying in the present invention.
- Figure 18 shows a magnified image generated with the feedback method.
- Same numerals in Figures 1-18 are assigned to similar elements in all the figures. Embodiments of the invention are discussed below with reference to Figures 1-18. However, those skilled in the art will readily appreciate that the detailed description given herein with respect to these figures is for explanatory purposes as the invention extends beyond these limited embodiments.
- Figure 4 shows one
preferred method 100 to magnify a digital image with many original pixels. The pixels are arranged with a certain resolution, for example, 300 dots or pixels per inch. Each pixel has a level, which describes the intensity at that pixel. For a continuous tone image, the intensity level may vary from 0 to 255. For a non-continuous tone image, such as a halftone image, preferably, it is transformed to a continuous tone image before it is magnified; this process will be further described later in the detailed description. - In the image, any region with a crisp-edge is a region with a significant change in level. The first step in the
preferred embodiment 100 is to analyze, 102, the image to pick all the locations with significant change in levels. This can be done visually or analytically. - One can visually identify all the regions with sharp changes in intensity. However, this can be a tedious task.
- Figure 5 shows one preferred analytical method, 200, to pick locations with significant changes in levels. Each pixel is represented, 202, by an impulse function with its magnitude being the level at that pixel. Then select, 204, a singly-peaked continuous function. As an example, this can be a normalized Gaussian function, having a second standard deviation covering the width of 3 original pixels. With the image being two-dimensional, the selected function is a two-dimensional surface. The selected continuous function is differentiated, 206, twice. If the function is a Gaussian function, then the second derivative is known as the Laplacian of the Gaussian. The result from the differentiation step is convolved, 208, with all the impulse functions, and the outputs are superimposed spatially to generate a convolved result. Pick, 212, all the locations that have substantially zero value from the convolved result. All the crisp-edge locations in the image before magnification correspond to locations in the convolved results with substantially zero values.
- Figure 6 shows one preferred method, 250, to pick locations that have substantially zero values or to pick the "zero-crossings" from the convolved results. There are a few ways to select the "zero-crossings." One can solve analytically the convolved result to find all the locations having zeros. This calculation might be quite cumbersome. Another preferred way is to solve the convolved result at selected positions, as shown in Figure 6, in combination with Figure 7.
- Assume that one wants to magnify the image N times, and the minimum distance between adjacent original pixels is D. In the present example shown in Figure 7, N is 2, and D is 3.3 mils for an image with a 300 dots/inch resolution. In picking, 250, locations with substantially zero value, first, one forms, 252, temporary pixels at positions offset from the original pixels, with the minimum distance between adjacent temporary pixels being D/N. In Figure 7, the crosses represent the locations of the original pixels at 300 dots per inch; the circles represent the temporary pixels; and the line marked EDGE represents the theoretical locations with significant changes in levels. The minimum distance between adjacent circles in Figure 7 is D/N or 1.67 mils, which gives a resolution of 600 dots per inch. The circles are offset from the crosses. The amount of offset is D/(2N) in both the X and the Y direction, with N being 2.
- The convolved result is evaluated, 254, at the temporary pixels. The levels of adjacent temporary pixels are compared, 256, to select, 258, those pairs of adjacent temporary pixels whose levels change sign. For example, there will be a sign change between the levels at
circles circles - From every selected pair, choose, 262, the location of one temporary pixel whose level is closer to zero. Refer back to Figure 7, for the
pair temporary pixel 276 is selected. - The
method 250 may be used in combination with Figure 8 as another scheme to pick locations that have substantially zero values or to pick the "zero-crossings" from the convolved result. This scheme reduces the amount of computation required and is especially suitable for even number magnification higher than 2. First, one forms, 252, not just temporary pixels, whose positions are shown as circles, but also target pixels, whose positions are shown as triangles. The locations of the temporary pixels have a resolution doubling the original resolution, with some of them not displaced from the original pixels. The locations of the target pixels have a resolution four times the original resolution, and are displaced from the temporary pixels by D/(2N) in both the X and the Y direction, with N being 4 for the example shown in Figure 8. The levels are again evaluated, 254, at the temporary pixels, with adjacent temporary pixels compared, 256, to select, 258, those pairs with sign change in their levels. - From every selected pair of temporary pixels, choose the location of an adjacent target pixel. One choosing process is as follows: for a selected pair of temporary pixels along the X-direction with a certain Y coordinate, consider the pair of the target pixels with a smaller Y coordinate; and for a selected pair along the Y-direction with a certain X coordinate, consider the pair of target pixels with a larger X coordinate. For the target pixels considered, choose the location of the target pixel that is closer to one of the temporary pixel; that temporary pixel has a level that is closer to zero than the other temporary pixel. If both temporary pixels have the same level, choose either one of the target pixels. This scheme reduces the amount of computation required as compared to the scheme described in Figure 7.
- As an example, in the vicinity of the
original pixels temporary pixels target pixels temporary pixels target pixel 288 being physically closer to thetemporary pixel 283, thetarget pixel 288 is chosen. - Figure 9 shows another
preferred embodiment 300 to pick locations with significant changes in levels. Each pixel is again represented, 302, by an impulse function with its magnitude being the level at that pixel. Then again select, 304, a singly-peaked continuous function, such as the same Gaussian function picked in theprevious embodiment 200. The selected function is convolved, 306, with all the impulse functions, and the outputs are superimposed together to get a convolved result. - The convolved result is differentiated, 310, twice to get a differentiated result. After the differentiation step, pick, 312, all the locations that have substantially zero from the differentiated result. All the crisp-edge locations have values that are substantially zero.
- Again one can use the method in Figure 6 in combination with Figure 7 or Figure 8 to pick locations that have substantially zero values or the "zero-crossings" from the differentiated result. The main difference is at the
step 254. Instead of evaluating the convolved result, one evaluates the differentiated result. - Due to noise in any system, there might be locations giving substantially zero values by the convolution step in the
embodiment 200 or the differentiation step in theembodiment 300 that are not crisp-edge regions. Thus, it is preferred to threshold, 214 or 314, the picked locations. Thresholding, 214 or 314, means removing some of the picked locations if the local level variances of the picked locations are less than a pre-selected value. One can calculate local variances by finding the slopes or changes of the levels in the picked locations. If the slope is less than a certain pre-selected value, the picked locations are dropped. The pre-selected value should be small because it is acceptable to have more locations which are perceived to be zero-crossings. However, missing zero-crossings would not be preferable. In one example, the pre-selected value is 0.03. - Refer back to Figure 4. After the locations with significant change in levels are picked, 102, the locations should be connected, 104, to form boundaries for an edge map. However not all the locations should be connected. In one preferred embodiment, one connects any two picked locations to form a boundary if the distance between the locations is less than or equal to 1.414 (or the square root of 2) times the minimum distance between two adjacent original pixels. For a digital image with 300 dots per inch, the minimum distance is 3.3 mils, and 1.414 times of 3.3 is equal to 4.7 mils. Figure 10 shows an example of a two-dimensional edge map formed by the above described process on the original image shown in Figure 1. That map is enlarged to the size of the desired magnified image.
- After the edge map is formed, it is projected, 106, onto the digital image. Then one generates, 108, many additional pixels in locations among the original pixels by manipulating the original pixels without crossing any boundaries set by the edge map. The number of additional pixels formed depends on the magnification desired. If one wants M times magnification in one dimension and N times in another dimension, one needs to increase the number of pixels in the image to M times N, which means forming additional number of pixels in appropriate locations.
- Two different examples would be used to illustrate the methods to generate additional pixels. Both of them depend on interpolating the levels of one or more original pixels.
- Figures 11A-D illustrate the first example, which is especially suitable for even numbers of magnifications. Crosses represent the locations of original pixels, for example, 379 and 383. Triangles represent the locations of additional pixels, for example, 385. In calculating the level of the
additional pixel 385, one interpolates the levels of the original pixels, 377, 379, 381; but one does not use the level of theoriginal pixel 383 because that would imply crossing a boundary,EDGE 1, in the edge map. - Numerous methods may be used to find the levels at the additional pixels. One method is to use bilinear interpolation, which is well known to those with ordinary skill in the art, and will not be further described. Figures 11B-D show a few preferred methods for finding the levels at the additional pixels using the edge map.
- Figure 11B shows an example of finding the level at
additional pixel 385, based on the levels at threeoriginal pixels fictitious pixel 382. Thefictitious pixel 382 is located at the location of theoriginal pixel 383. From the levels ofpixels additional pixel 385 using bilinear interpolation. The above calculations use proportionality based on distance and will not be further discussed here. - Figure 11C shows examples of finding the level at the additional pixel 392, based on two original pixels, 383, 387. Two situations will be described, the first situation involving just one boundary next to the original pixels, and the second situation involving the original pixels surrounded by two boundaries. In the first situation, with only EDGE2, but without EDGE3, one preferred method is to extrapolate based on the neighbors of the original pixels. The value at the fictitious pixel 390 is extrapolated from the levels in the
original pixels fictitious pixel 391 is extrapolated from the levels in theoriginal pixels original pixels fictitious pixels 390 and 391 respectively. The locations of thefictitious pixels 390 and 391 are the locations of theoriginal pixels pixels - Figure 11D shows an example of finding the level at the
additional pixel 394, based on one original pixel, 381. One method is to duplicate the level of the original pixel, 381, at the locations of the fictitious pixels, 393, 395 and 397, which are the locations of theoriginal pixels pixels additional pixel 394 using bilinear interpolation. - Although the above calculations are based on bilinear interpolation, other interpolation techniques may be used, such as other higher order interpolation schemes.
- Figure 11E illustrates the second example to find the levels at additional levels, which is especially suitable for odd numbers of magnifications. The main difference between this example and the first example is that some additional pixels are located at the original pixel locations. The methods to generate the levels of the additional pixels can be similar to the methods as described in Figures 11B-D.
- In another preferred embodiment, there is an additional step to improve the magnified image. Suppose that the actual boundary where the levels of the image change significantly is at a small distance away from the boundary found. Then the level of some additional pixels that are in close proximity to the found boundary may have values quite different from its immediate neighboring pixels. This may be due to the blurring effect of the detectors measuring the original image. Such abnormalities can be corrected by the
methods 450 shown in Figure 12A, in combination with a graphical representation shown in Figure 12B. First, a size of awindow 475 is selected, 452. One example is a size that covers 2 by 2 original pixels. Then center, 454, the window at every additional pixel close to a found boundary. For example, center thewindow 475 at theadditional pixel 477. With the window centered, one calculates, 456, the average level of the original pixels on the side of the boundary of the additional pixel within the window. For example, calculate the average level of the original pixels, 479, 481 and 483. If the average level differs by more than a pre-set value from the level of the additional pixel, replace, 458, the level of the additional pixel with the average value. In one embodiment, the pre-set value is one standard deviation of the levels of the original pixels. This process is performed on every pixel in close proximity to a boundary. If an additional pixel is bounded by two boundaries, then the original pixels averaged are only those within the window on the side of the additional pixel and within the two boundaries. - After the formation of all the additional pixels, the distances between the pixels are expanded, 112, so that the resolution of the locations of the pixels are the same as the original pixels. For example, all the crosses in the digital image in Figure 11A has a resolution of 300 pixels per inch. After the formation of the additional pixels, the triangles have a resolution of 600 pixels per inch. The size of the image is then expanded by expanding the distances between the additional pixels so that the resolution of the locations of the additional pixels goes back to 300 pixels per inch. This means expanding the image two times in each dimension.
- Figure 13 shows the magnified digital image based on the above described method. With the edge map, the magnified image is more crisp than the image in Figure 3, for example, as in the stamens of the off-white lily on the right hand side of the Figure or as in the gaps between flowers.
- In the above description, the distances between pixels are expanded at the end. Figure 14 shows a modified
preferred embodiment 498, in which the distances between pixels are expanded in the beginning. First, the original digital image is enlarged, 490 by expanding proportionally the distances between the original pixels. Then an edge map is produced, 492, by one of the preferred methods described above. The produced edge map is projected, 494, onto the enlarged image to generate, 496, additional pixels by methods as described above. After the step of generating, the resolution of the locations of the pixels becomes the same as the original pixels, and the digital image is magnified. - In another preferred embodiment, the image in Figure 13 is further improved by a feedback method, 500, shown in Figure 15. In this feedback method, the original digital image is duplicated, 502, to form a duplicated image. The duplicated image is then magnified, 504, R times. This magnification process may be based on the edge map method as described above or may be based on other means. However, the magnification process is preferably not the pixel replication process, which generates the image shown in Figure 2. The magnified image is then de-magnified, 506, R times.
- The demagnification process is based on averaging. Every fixed number of pixels are replaced with one pixel whose level is the average level of the fixed number of pixels. Figure 16 shows a two-dimensional example of the averaging process for a two-times demagnification. Every four pixel is replaced by one pixel; for example, the
pixels image 570 is replaced by thepixel 558 in thede-magnified image 572. The level of thepixel 558 is equal to the average of the levels of the four pixels in the magnifiedimage 570. The location of thepixel 558 is in the center of the fourpixels - Then, one calculates, 508, the difference between the de-magnified image and the original digital image at every pixel. This is a pixel-to-pixel comparison method. Based on the differences found, the duplicated image is modified, 512. This modifying step includes modifying every pixel of the duplicated image by a selected multiple of the difference found at that pixel from the step of calculation, 508. Figure 16 shows an example of the step of modifying. The original image has a
pixel 600 with a grey-level of 200. That pixel is compared to the pixel, 602, at the corresponding position in the de-magnified image. Thepixel 602 has a grey-level of 240. The difference value is 40. A multiple of the difference value is used to modify the corresponding pixel, 604, in the duplicated image. If the multiple is too large, one may get an "oscillation." If the multiple is too small, one may take a long time before one gets satisfactory results. The multiple value can be found experimentally. Note also that the three pixels, 600, 602 and 604, are all in substantially the same location for the pixel-to-pixel operation described. - After the modification of the duplicated image, one checks the differences found in the step of calculation, 508. If the average of the absolute of the differences found is bigger than a pre-selected value, one repeats, 514, from the step of magnifying, 504, the duplicated image R times. This pre-selected value depends on the accuracy one desires. One is not limited to using the absolute values of the differences. Other values may be used, such as the square of the differences. Anyway, if the differences found are less than the pre-selected value, the feedback process stops 516, and the magnified duplicated image becomes the final result of magnification.
- Figure 18 shows an example of applying the feedback method on the image shown in Figure 13, with the pre-selected value set as 2.5 and the multiple value set as 1. The multiple value preferably should be between 0 and 2. The magnified image formed after the feedback method has significantly better level consistency. Also, many details in the original image are maintained in the magnified image; for example, compare the veins on the petals of the flowers in Figure 13 and Figure 17.
- This feedback process is applicable not just to two-dimensional images. It is equally applicable for magnifying images with other dimensions, such as one or three dimensional images. The difference is the fixed number of pixels to be averaged. For one-dimensional magnification of R times, the fixed number is R. For three dimensional magnification of R times, the fixed number is R * R * R. The above preferred embodiment describes the magnification factors being equal for all the dimensions. In other preferred embodiments, the magnification factors are different for different dimensions. In other words, for example, the two dimensional magnification may be R * N, with R not equal to N.
- The present invention is not just applicable to continuous tone images. It can also be applied to non-continuous tone images, such as halftone images. As an example, for halftone images, the images are first changed to continuous tones, based on inverse halftoning techniques. Then images are magnified as described above. After the magnification process, the images are halftoned to produce halftone images. Inverse halftoning and halftoning techniques are obvious to those with ordinary skill in the arts and will not be further described in the present disclosure.
- From the foregoing it should be appreciated that an edge map significantly maintains the crisp-edge features of a magnified image, and the feedback method further improves the quality of the magnified image. The invention is applicable to many other areas than those just described. For example, the invention is applicable to three-dimensional images, such as a holographic image; in that case, the edge map would not be just lines, but surfaces. Another example is to apply the invention to color images. For monochrome images, the level of each pixel is the intensity at that pixel. For color images, the level of each pixel can be identified by the intensities of their primaries, such as red, green and blue; the present invention is applied to each color individually, and then the results of the three colors are combined at each pixel to re-generate the color images. Another preferred embodiment for color images is to identify the level of each pixel as its luminous at that pixel. Then the present invention is applied to the luminous value as described above.
- This invention can be applied to one-dimensional images. In that case, the image is a one-dimensional waveform or a signal, like a voice signal. Another area of application is in video. In that case, one magnifies two-dimensional images, frame by frame, leaving the third dimension, which is time, alone.
- Other embodiments of the invention will be apparent to the skilled in the art from a consideration of this specification or practice of the invention disclosed herein. It is intended that the specification and examples be considered as exemplary only, with the true scope and spirit of the invention being indicated by the following claims.
Claims (12)
- A method (100, 498) of magnifying a digital image with a plurality of original pixels, each pixel having a level, the locations of the pixels having a resolution, the method comprising the steps of:
producing (102, 104, 492) an edge map with a plurality of boundaries from the digital image;
projecting (106, 492) the edge map onto an image with the same pixels and the same number of pixels as the digital image; and
generating (108, 496) one or more additional pixels in locations among the original pixels by manipulating the levels of one or more original pixels without crossing any boundaries set by the edge map;
such that the total number of pixels is related to the number of original pixels and the magnification. - A method (100) as recited in claim 1 wherein:
the step of projecting (106) includes projecting onto the original image; and
the method further comprises the step of:
expanding (112), after the step of generating (108), the distances between the pixels so that the resolution of the locations of the pixels becomes the same as the original pixels. - A method (498) as recited in claim 1
further comprising the step of:
enlarging (490), before producing (492) the edge map, the original image to the size of the magnified image by expanding proportionally the distances separating the original pixels; and
wherein:
the step of projecting (494) includes projecting onto the enlarged image; and
after the step of generating (496), the resolution of the locations of the pixels becomes the same as the original pixels. - A method as recited in Claims 2 or 3, wherein the step of producing further comprises the steps of:
analyzing (102) the digital image to pick locations in neighborhoods where levels change significantly; and
connecting (104) any two picked locations to form an edge if the distance between the locations is less than or equal to the minimum distance between two adjacent original pixels. - A method as recited in Claim 4 wherein the step of analyzing (102) further comprises:
representing (202) each pixel by an impulse function with its magnitude being equal to the level of its corresponding pixel;
selecting (204) a singly-peaked continuous function;
differentiating (206) the selected function twice;
convolving (208) the differentiated function with the impulse functions; and
picking (212) the locations where the convolving step generates substantially a zero. - A method as recited in Claim 4 wherein the step of analyzing (102) further comprising:
representing (302) each pixel by an impulse function with its magnitude being equal to the level of its corresponding pixel;
selecting (304) a singly-peaked continuous function;
convolving (306) the selected function with the impulse functions;
differentiating (310) twice the convolved results; and
picking (312) the locations where the differentiating step generates substantially a zero. - A method as recited in Claims 5 or 6, wherein the step of picking (212, 312) further comprises the step of thresholding (214, 314), which removes any picked location with its local level variances less than a pre-selected value.
- A method as recited in Claim 5 wherein the step of picking (212) further comprises the steps of:
forming (252) temporary pixels;
evaluating (254) the levels at the temporary pixels from the results of the convolving step;
comparing (256) the levels of adjacent temporary pixels;
selecting (258) those pairs of adjacent temporary pixels whose levels change sign; and
choosing (262) the location based on the temporary pixel, from every selected pair, whose level is closer to zero. - A method as recited in Claim 6 wherein the step of picking (312) further comprises the steps of:
forming (252) temporary pixels;
evaluating (254) the levels at the temporary pixels from the results of the differentiating step;
comparing (256) the levels of adjacent temporary pixels;
selecting (258) those pairs of adjacent temporary pixels whose levels change sign; and
choosing (262) the location based on the temporary pixel, from every selected pair, whose level is closer to zero. - A method as recited in Claims 2 or 3, wherein the step of generating (108, 496) further comprises the step of interpolating the levels of one or more original pixels to create the levels of one or more additional pixels in the immediate neighborhood of the original pixels.
- A method as recited in Claims 2 or 3, wherein the step of generating (108, 496) further comprises the steps of:
selecting (452) a window size; and
for every additional pixel in the immediate neighborhood of the boundaries,
centering (454) the window to that additional pixel,
calculating (456) the average level of the original pixels within the window and on the side of the boundary of the additional pixel, and
replacing (458) the level of the additional pixel with the average level if the level of the additional pixel differs by more than a certain value from the average. - A method (100, 498) as recited in claim 1 wherein the image is a halftone image, and the method further comprising two additional steps,
inverse halftoning the image to form a continuous image before performing the method as recited in claim 1; and
halftoning the resultant image to form a halftone image after performing the method as recited in claim 1.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP00102982A EP1018705B1 (en) | 1994-04-14 | 1994-12-15 | Magnifying digital image using edge mapping |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US08/227,764 US5446804A (en) | 1994-04-14 | 1994-04-14 | Magnifying digital image using edge mapping |
US227764 | 1994-04-14 |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP00102982A Division EP1018705B1 (en) | 1994-04-14 | 1994-12-15 | Magnifying digital image using edge mapping |
Publications (3)
Publication Number | Publication Date |
---|---|
EP0677820A2 true EP0677820A2 (en) | 1995-10-18 |
EP0677820A3 EP0677820A3 (en) | 1996-02-28 |
EP0677820B1 EP0677820B1 (en) | 2002-03-20 |
Family
ID=22854363
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP94119855A Expired - Lifetime EP0677820B1 (en) | 1994-04-14 | 1994-12-15 | Magnifying digital image using edge mapping |
EP00102982A Expired - Lifetime EP1018705B1 (en) | 1994-04-14 | 1994-12-15 | Magnifying digital image using edge mapping |
Family Applications After (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP00102982A Expired - Lifetime EP1018705B1 (en) | 1994-04-14 | 1994-12-15 | Magnifying digital image using edge mapping |
Country Status (4)
Country | Link |
---|---|
US (2) | US5446804A (en) |
EP (2) | EP0677820B1 (en) |
JP (1) | JP3785202B2 (en) |
DE (2) | DE69430336T2 (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0803841A2 (en) * | 1996-04-25 | 1997-10-29 | Canon Kabushiki Kaisha | Image processing apparatus and method |
EP1227436A2 (en) * | 2001-01-17 | 2002-07-31 | Eastman Kodak Company | Method for zooming digital images |
WO2013106266A1 (en) * | 2012-01-10 | 2013-07-18 | Apple Inc. | Super-resolution image using selected edge pixels |
WO2014173970A1 (en) * | 2013-04-26 | 2014-10-30 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Upsampling and signal enhancement |
CN104239909A (en) * | 2014-08-11 | 2014-12-24 | 北京捷通华声语音技术有限公司 | Method and device for recognizing images |
WO2016137725A1 (en) * | 2015-02-25 | 2016-09-01 | Qualcomm Incorporated | Digital zoom methods and systems |
WO2022142080A1 (en) * | 2020-12-30 | 2022-07-07 | 稿定(厦门)科技有限公司 | Image comparison method and apparatus |
Families Citing this family (57)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5446804A (en) * | 1994-04-14 | 1995-08-29 | Hewlett-Packard Company | Magnifying digital image using edge mapping |
JP2973899B2 (en) * | 1995-11-17 | 1999-11-08 | 富士ゼロックス株式会社 | Image processing device |
US5914725A (en) * | 1996-03-07 | 1999-06-22 | Powertv, Inc. | Interpolation of pixel values and alpha values in a computer graphics display device |
JPH09326958A (en) * | 1996-06-05 | 1997-12-16 | Sony Corp | Image processing unit and processing method |
KR100283574B1 (en) * | 1996-08-27 | 2001-03-02 | 윤종용 | Monitor screen size control circuit and its control method |
US5982373A (en) * | 1996-11-12 | 1999-11-09 | Chromatic Research, Inc. | Dynamic enhancement/reduction of graphical image data resolution |
US6075926A (en) * | 1997-04-21 | 2000-06-13 | Hewlett-Packard Company | Computerized method for improving data resolution |
US5912683A (en) * | 1997-08-25 | 1999-06-15 | Lexmark International, Inc. | Method of printing with an ink jet printer using an enhanced horizontal resolution |
US6707572B1 (en) * | 1997-09-12 | 2004-03-16 | Tesseron, Ltd. | Filtering method to reduce pixel density |
US6111583A (en) * | 1997-09-29 | 2000-08-29 | Skyline Software Systems Ltd. | Apparatus and method for three-dimensional terrain rendering |
US6639593B1 (en) * | 1998-07-31 | 2003-10-28 | Adobe Systems, Incorporated | Converting bitmap objects to polygons |
US6219465B1 (en) | 1998-09-23 | 2001-04-17 | Xerox Corporation | High quality digital scaling using pixel window averaging and linear interpolation |
US6476873B1 (en) | 1998-10-23 | 2002-11-05 | Vtel Corporation | Enhancement of a selectable region of video |
US6546117B1 (en) * | 1999-06-10 | 2003-04-08 | University Of Washington | Video object segmentation using active contour modelling with global relaxation |
AUPP779798A0 (en) * | 1998-12-18 | 1999-01-21 | Canon Kabushiki Kaisha | A modified kernel for image interpolation |
AUPP779898A0 (en) | 1998-12-18 | 1999-01-21 | Canon Kabushiki Kaisha | A method of kernel selection for image interpolation |
US20030158786A1 (en) * | 1999-02-26 | 2003-08-21 | Skyline Software Systems, Inc. | Sending three-dimensional images over a network |
US6424749B1 (en) * | 1999-03-30 | 2002-07-23 | Matsushita Electric Industrial Co., Ltd. | System and method for scaling combined video and computer generated imagery |
US6366292B1 (en) | 1999-06-22 | 2002-04-02 | Oak Technology, Inc. | Scaling method and apparatus for a flat panel display |
AUPQ377899A0 (en) | 1999-10-29 | 1999-11-25 | Canon Kabushiki Kaisha | Phase three kernel selection |
AUPQ377599A0 (en) | 1999-10-29 | 1999-11-25 | Canon Kabushiki Kaisha | Colour clamping |
US6718074B1 (en) * | 2000-06-02 | 2004-04-06 | Cognex Corporation | Method and apparatus for inspection for under-resolved features in digital images |
US6650790B1 (en) * | 2000-06-09 | 2003-11-18 | Nothshore Laboratories, Inc. | Digital processing apparatus for variable image-size enlargement with high-frequency bandwidth synthesis |
US6868186B1 (en) * | 2000-07-13 | 2005-03-15 | Ceva D.S.P. Ltd. | Visual lossless image compression |
US6717622B2 (en) * | 2001-03-30 | 2004-04-06 | Koninklijke Philips Electronics N.V. | System and method for scalable resolution enhancement of a video image |
US7053953B2 (en) * | 2001-12-21 | 2006-05-30 | Eastman Kodak Company | Method and camera system for blurring portions of a verification image to show out of focus areas in a captured archival image |
US7543326B2 (en) * | 2002-06-10 | 2009-06-02 | Microsoft Corporation | Dynamic rate control |
US20030235250A1 (en) * | 2002-06-24 | 2003-12-25 | Ankur Varma | Video deblocking |
US7119837B2 (en) * | 2002-06-28 | 2006-10-10 | Microsoft Corporation | Video processing system and method for automatic enhancement of digital video |
US6907194B2 (en) | 2002-11-12 | 2005-06-14 | Eastman Kodak Company | Camera having continuously cropping viewfinder |
US7561793B2 (en) * | 2002-11-12 | 2009-07-14 | Eastman Kodak Company | User interface for controlling cropping in electronic camera |
US7006764B2 (en) * | 2002-11-12 | 2006-02-28 | Eastman Kodak Company | User interface for controlling cropping in electronic camera |
US7327890B2 (en) | 2002-12-20 | 2008-02-05 | Eastman Kodak Company | Imaging method and system for determining an area of importance in an archival image |
JP3997415B2 (en) * | 2002-12-25 | 2007-10-24 | セイコーエプソン株式会社 | Edge generation apparatus, edge generation method, and edge generation program |
JP3952188B2 (en) * | 2003-01-08 | 2007-08-01 | セイコーエプソン株式会社 | Image interpolation apparatus, image interpolation method, and image interpolation program |
JP2005012740A (en) * | 2003-04-21 | 2005-01-13 | Toshiba Corp | Image processor and image processing method |
JP2007527567A (en) * | 2003-07-02 | 2007-09-27 | セラーテム・テクノロジー・インコーポレイテッド | Image sharpening with region edge sharpness correction |
US20050036711A1 (en) * | 2003-08-14 | 2005-02-17 | Fuji Photo Film Co., Ltd. | Image interpolation apparatus and method, and edge detecting apparatus and method |
US7764839B2 (en) * | 2003-08-14 | 2010-07-27 | Fujifilm Corporation | Edge detecting apparatus and method, and image size enlarging and reducing apparatus and method |
US7248752B2 (en) * | 2003-08-27 | 2007-07-24 | Electronics For Imaging, Inc. | Methods and apparatus for converting the resolution of binary image data |
KR100519776B1 (en) * | 2003-11-24 | 2005-10-07 | 삼성전자주식회사 | Method and apparatus for converting resolution of video signal |
US20050134719A1 (en) * | 2003-12-23 | 2005-06-23 | Eastman Kodak Company | Display device with automatic area of importance display |
EP1721458A1 (en) * | 2004-02-23 | 2006-11-15 | Koninklijke Philips Electronics N.V. | Reducing artefacts in scan-rate conversion of image signals by combining interpolation and extrapolation of images |
US8659619B2 (en) | 2004-03-26 | 2014-02-25 | Intellectual Ventures Fund 83 Llc | Display device and method for determining an area of importance in an original image |
KR100562937B1 (en) * | 2004-08-11 | 2006-03-22 | 엘지전자 주식회사 | Method and Apparatus for Processing Image of Display Device |
US7545391B2 (en) * | 2004-07-30 | 2009-06-09 | Algolith Inc. | Content adaptive resizer |
KR100648308B1 (en) * | 2004-08-12 | 2006-11-23 | 삼성전자주식회사 | Resolution conversion method and apparatus |
TWI312633B (en) * | 2005-06-02 | 2009-07-21 | Ind Tech Res Inst | A composite method and apparatus for scaling digital image |
TWI320914B (en) * | 2006-07-28 | 2010-02-21 | Via Tech Inc | Weight-adjusted apparatus and method thereof |
US20100260435A1 (en) | 2007-12-21 | 2010-10-14 | Orlick Christopher J | Edge Directed Image Processing |
US8116532B2 (en) * | 2008-08-15 | 2012-02-14 | Baker Hughes Incorporated | Extraction of processed borehole image elements to create a combined image |
KR101548285B1 (en) * | 2009-01-20 | 2015-08-31 | 삼성전자주식회사 | Apparatus and method for obtaining high resolution image |
TWI412730B (en) | 2009-06-08 | 2013-10-21 | Wistron Corp | Methods and device for detecting distance, identifying positions of targets, and identifying a current position in a smart portable device |
US9129409B2 (en) * | 2009-07-29 | 2015-09-08 | Qualcomm Incorporated | System and method of compressing video content |
US8391647B1 (en) | 2010-02-17 | 2013-03-05 | Hewlett-Packard Development Company, L.P. | Pixel replacement |
JP5810631B2 (en) * | 2011-05-26 | 2015-11-11 | コニカミノルタ株式会社 | Ultrasonic diagnostic equipment |
CN107391843B (en) * | 2017-07-21 | 2020-12-18 | 深圳市置辰海信科技有限公司 | Dynamic calculation method for submarine cable operation cost |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5054100A (en) * | 1989-11-16 | 1991-10-01 | Eastman Kodak Company | Pixel interpolator with edge sharpening |
US5131057A (en) * | 1990-02-12 | 1992-07-14 | Wright State University | Method for video-to-printing image resolution conversion |
EP0645736A2 (en) * | 1993-09-27 | 1995-03-29 | Canon Kabushiki Kaisha | Image processing apparatus |
Family Cites Families (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4611349A (en) * | 1984-08-13 | 1986-09-09 | Xerox Corporation | Halftone image scaling |
JPH0769958B2 (en) * | 1988-09-20 | 1995-07-31 | 沖電気工業株式会社 | Image conversion processing method |
NL8900284A (en) * | 1989-02-06 | 1990-09-03 | Oce Nederland Bv | METHOD AND APPARATUS FOR CONVERTING THE RESOLUTION OF AN IMAGE DISPLAYED BY A PATTERN OF TWO-VALUE PIXELS |
US5305398A (en) * | 1989-10-10 | 1994-04-19 | Unisys Corporation | Method and apparatus for scaling image data |
US5142592A (en) * | 1990-12-17 | 1992-08-25 | Moler Keith E | Method and apparatus for detection of parallel edges in image processing |
JP2619758B2 (en) * | 1991-12-20 | 1997-06-11 | 大日本スクリーン製造株式会社 | Image data compression method |
JP2619759B2 (en) * | 1991-12-20 | 1997-06-11 | 大日本スクリーン製造株式会社 | Image data compression method |
US5446804A (en) * | 1994-04-14 | 1995-08-29 | Hewlett-Packard Company | Magnifying digital image using edge mapping |
-
1994
- 1994-04-14 US US08/227,764 patent/US5446804A/en not_active Expired - Lifetime
- 1994-12-15 EP EP94119855A patent/EP0677820B1/en not_active Expired - Lifetime
- 1994-12-15 DE DE69430336T patent/DE69430336T2/en not_active Expired - Fee Related
- 1994-12-15 EP EP00102982A patent/EP1018705B1/en not_active Expired - Lifetime
- 1994-12-15 DE DE69430191T patent/DE69430191T2/en not_active Expired - Fee Related
-
1995
- 1995-03-29 US US08/412,640 patent/US5661824A/en not_active Expired - Lifetime
- 1995-04-13 JP JP11251695A patent/JP3785202B2/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5054100A (en) * | 1989-11-16 | 1991-10-01 | Eastman Kodak Company | Pixel interpolator with edge sharpening |
US5131057A (en) * | 1990-02-12 | 1992-07-14 | Wright State University | Method for video-to-printing image resolution conversion |
EP0645736A2 (en) * | 1993-09-27 | 1995-03-29 | Canon Kabushiki Kaisha | Image processing apparatus |
Non-Patent Citations (3)
Title |
---|
PROCEEDINGS OF THE 4TH INTERNATIONAL JOINT CONFERENCE ON PATTERN RECOGNITION, KYOTO, JAPAN, 7-10 NOV. 1978, 1978, KYOTO, JAPAN, KYOTO, UNIV, JAPAN, pages 946-950, TRIENDL E E 'How to get the edge into the map (remote sensing)' * |
ROSENFELD A., KAK A. 'Digital Picture Processing' 1982 , ACADEMIC PRESS , ORLANDO, FLORIDA * page 89, paragraph 2 - page 91, paragraph 4 * * |
THIRD INTERNATIONAL CONFERENCE ON IMAGE PROCESSING AND ITS APPLICATIONS (CONF. PUBL. NO.307), WARWICK, UK, 18-20 JULY 1989, 1989, LONDON, UK, IEE, UK, pages 664-668, ATWOOD G H ET AL 'Image expansion using interpolation and heuristic edge following' * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0803841A2 (en) * | 1996-04-25 | 1997-10-29 | Canon Kabushiki Kaisha | Image processing apparatus and method |
EP0803841A3 (en) * | 1996-04-25 | 1998-12-23 | Canon Kabushiki Kaisha | Image processing apparatus and method |
US6009213A (en) * | 1996-04-25 | 1999-12-28 | Canon Kabushiki Kaisha | Image processing apparatus and method |
EP1227436A2 (en) * | 2001-01-17 | 2002-07-31 | Eastman Kodak Company | Method for zooming digital images |
EP1227436A3 (en) * | 2001-01-17 | 2005-03-09 | Eastman Kodak Company | Method for zooming digital images |
WO2013106266A1 (en) * | 2012-01-10 | 2013-07-18 | Apple Inc. | Super-resolution image using selected edge pixels |
WO2014173970A1 (en) * | 2013-04-26 | 2014-10-30 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Upsampling and signal enhancement |
US9830682B2 (en) | 2013-04-26 | 2017-11-28 | Fraunhofer-Gesellschaft Zur Foerderung Der Angewandten Forschung E.V. | Upsampling and signal enhancement |
CN104239909A (en) * | 2014-08-11 | 2014-12-24 | 北京捷通华声语音技术有限公司 | Method and device for recognizing images |
CN104239909B (en) * | 2014-08-11 | 2018-03-16 | 北京捷通华声语音技术有限公司 | A kind of recognition methods of image and device |
WO2016137725A1 (en) * | 2015-02-25 | 2016-09-01 | Qualcomm Incorporated | Digital zoom methods and systems |
WO2022142080A1 (en) * | 2020-12-30 | 2022-07-07 | 稿定(厦门)科技有限公司 | Image comparison method and apparatus |
Also Published As
Publication number | Publication date |
---|---|
US5446804A (en) | 1995-08-29 |
EP1018705A3 (en) | 2000-07-26 |
EP1018705A2 (en) | 2000-07-12 |
US5661824A (en) | 1997-08-26 |
JP3785202B2 (en) | 2006-06-14 |
JPH07302333A (en) | 1995-11-14 |
EP0677820A3 (en) | 1996-02-28 |
DE69430191D1 (en) | 2002-04-25 |
DE69430336D1 (en) | 2002-05-08 |
DE69430191T2 (en) | 2003-08-14 |
EP0677820B1 (en) | 2002-03-20 |
DE69430336T2 (en) | 2002-08-08 |
EP1018705B1 (en) | 2002-04-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US5446804A (en) | Magnifying digital image using edge mapping | |
EP0677821A2 (en) | Magnify a digital image using feedback | |
EP2215597B1 (en) | Re-sizing image sequences | |
US6950211B2 (en) | Fine moire correction in images | |
CA2702165C (en) | Image generation method and apparatus, program therefor, and storage medium which stores the program | |
US6959117B2 (en) | Method and apparatus for deblurring and re-blurring image segments | |
US7142729B2 (en) | System and method of scaling images using adaptive nearest neighbor | |
US6980254B1 (en) | Image interpolation system and image interpolation method | |
JP2003143399A (en) | Image processing apparatus and program | |
WO2006090515A1 (en) | Image processing device, image processing method, and image display device | |
US7408559B2 (en) | Upscaling of anti-aliased graphical elements | |
JP2000242774A (en) | Image processing method and image processor | |
JP2011095861A (en) | Image processing apparatus, control method and program | |
CN111062867A (en) | Video super-resolution reconstruction method | |
US5828380A (en) | Method for processing images | |
JP3210264B2 (en) | Image processing apparatus and method | |
JP3026706B2 (en) | Image processing device | |
JP3092769B2 (en) | Image processing device | |
JP4569697B2 (en) | Image processing apparatus, image display apparatus, and image processing method | |
JP2002077622A (en) | Image processing apparatus and recording medium | |
JP4087010B2 (en) | Image processing device | |
JPH1063824A (en) | Interpolation smoothing device for picture data | |
JPH0383177A (en) | Dither image enlargement system | |
JPH08251397A (en) | Method and device for image processing | |
JP2009157915A (en) | Image processing apparatus, image processing method, and image display apparatus |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
AK | Designated contracting states |
Kind code of ref document: A2 Designated state(s): DE FR GB NL |
|
PUAL | Search report despatched |
Free format text: ORIGINAL CODE: 0009013 |
|
AK | Designated contracting states |
Kind code of ref document: A3 Designated state(s): DE FR GB NL |
|
17P | Request for examination filed |
Effective date: 19961112 |
|
17Q | First examination report despatched |
Effective date: 19990827 |
|
GRAG | Despatch of communication of intention to grant |
Free format text: ORIGINAL CODE: EPIDOS AGRA |
|
GRAG | Despatch of communication of intention to grant |
Free format text: ORIGINAL CODE: EPIDOS AGRA |
|
GRAG | Despatch of communication of intention to grant |
Free format text: ORIGINAL CODE: EPIDOS AGRA |
|
RAP1 | Party data changed (applicant data changed or rights of an application transferred) |
Owner name: HEWLETT-PACKARD COMPANY, A DELAWARE CORPORATION |
|
GRAG | Despatch of communication of intention to grant |
Free format text: ORIGINAL CODE: EPIDOS AGRA |
|
GRAH | Despatch of communication of intention to grant a patent |
Free format text: ORIGINAL CODE: EPIDOS IGRA |
|
GRAH | Despatch of communication of intention to grant a patent |
Free format text: ORIGINAL CODE: EPIDOS IGRA |
|
REG | Reference to a national code |
Ref country code: GB Ref legal event code: IF02 |
|
GRAA | (expected) grant |
Free format text: ORIGINAL CODE: 0009210 |
|
AK | Designated contracting states |
Kind code of ref document: B1 Designated state(s): DE FR GB NL |
|
REF | Corresponds to: |
Ref document number: 69430191 Country of ref document: DE Date of ref document: 20020425 |
|
ET | Fr: translation filed | ||
PLBE | No opposition filed within time limit |
Free format text: ORIGINAL CODE: 0009261 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: NO OPPOSITION FILED WITHIN TIME LIMIT |
|
26N | No opposition filed |
Effective date: 20021223 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: DE Payment date: 20060630 Year of fee payment: 12 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: DE Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20070703 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: NL Payment date: 20071223 Year of fee payment: 14 |
|
NLV4 | Nl: lapsed or anulled due to non-payment of the annual fee |
Effective date: 20090701 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: NL Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20090701 |
|
REG | Reference to a national code |
Ref country code: GB Ref legal event code: 732E Free format text: REGISTERED BETWEEN 20120329 AND 20120404 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: GB Payment date: 20121227 Year of fee payment: 19 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: FR Payment date: 20130110 Year of fee payment: 19 |
|
GBPC | Gb: european patent ceased through non-payment of renewal fee |
Effective date: 20131215 |
|
REG | Reference to a national code |
Ref country code: FR Ref legal event code: ST Effective date: 20140829 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: FR Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20131231 Ref country code: GB Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20131215 |